Ornn

The agent-facing skill-lifecycle API for AI agents.

Ornn official website — ornn.chrono-ai.fun

What is Ornn · How it works · SDK quickstart · Quickstart · Docs · Roadmap · Community · Contributing

What is Ornn

Ornn is an agent-facing skill-lifecycle API, not a human marketplace.

The primary consumer is the AI agent developer / agentic-system builder. Agents call Ornn directly — over HTTP or MCP — to manage their own skill lifecycle:

search → pull → install → execute → build → upload → share

Closest analog: npm registry + npm CLI fused, model-agnostic — works for Claude, GPT, Gemini, or any custom runtime. Not locked to a single model.

ornn-web is a secondary surface for skill owners and platform admins; it is not the primary product.

How it works

┌──────────────┐    HTTP / MCP    ┌──────────────┐    auth     ┌──────────┐
│   AI agent   │ ───────────────▶ │   ornn-api   │ ──────────▶ │  NyxID   │
│ (any model)  │                  │              │             └──────────┘
└──────────────┘                  │              │   storage   ┌──────────┐
       │                          │              │ ──────────▶ │ MongoDB  │
       │                          │              │             └──────────┘
       │                          │              │   sandbox   ┌──────────┐
       │ pull / execute           │              │ ──────────▶ │ OpenSbox │
       ▼                          └──────────────┘             └──────────┘
┌──────────────┐
│ Local skill  │
│   runtime    │
└──────────────┘

The agent talks to ornn-api through nyxid, which brokers authentication and authorization on the agent's behalf. Skills are versioned artifacts that the agent pulls, runs in a sandbox, and (optionally) publishes back.

For a deeper view, see docs/ARCHITECTURE.md.

SDK quickstart

Call Ornn directly from code. The SDKs wrap /api/v1/* and handle auth header injection, response-envelope unwrapping, and structured errors.

TypeScript (sdk/typescript)

# Pre-publish — install from this monorepo via your package manager's
# workspace / git subdirectory mechanism. See #473 for npm publish status.
npm install @chronoai/ornn-sdk

import { OrnnClient } from "@chronoai/ornn-sdk";

const ornn = new OrnnClient({
  baseUrl: "https://ornn.chrono-ai.fun",
  token: process.env.ORNN_TOKEN,
});
const result = await ornn.search({ q: "pdf parsing" });
console.log(result.items[0]);

Python (sdk/python)

pip install ornn-sdk

import os
from ornn_sdk import OrnnClient

ornn = OrnnClient(
    base_url="https://ornn.chrono-ai.fun",
    token=os.environ["ORNN_TOKEN"],
)
result = ornn.search(q="pdf parsing")
print(result.items[0])

Token sources are pluggable — for dynamic refresh flows, pass getToken (TS) / token_resolver (Python) instead of a static token. See sdk/typescript/README.md and sdk/python/README.md for the full reference.

Quickstart

Status: alpha. Surfaces and schemas can change before v1. Pin a release tag.

The shortest path to bringing an agent online with Ornn — no manual operator steps inside the agent loop:

1. Install the ChronoAI core service skill into your agent. This is the bootstrap skill — it introduces Ornn to the agent and drives the rest of setup.
2. Let the agent provision the NyxID CLI. On first run, the core skill instructs the agent to install nyxid. The agent follows the skill end-to-end.
3. Talk to the agent. Ask it to search, install, run, build, or publish skills. The agent learns the lifecycle through the same API it just connected to.

Once connected, an agent can hit the API directly. A minimal request shape (after nyxid is configured):

nyxid proxy request ornn-api GET /api/v1/skills?q=summarize

Full per-endpoint reference: ornn.chrono-ai.fun/docs.

Documentation

• Product docs — ornn.chrono-ai.fun/docs
• Architecture — docs/ARCHITECTURE.md
• Conventions — docs/CONVENTIONS.md
• Design system — docs/DESIGN.md

Roadmap

Tracked publicly on GitHub:

• Open issues & milestones — Issues · Milestones
• What shipped — Releases · per-package changelogs in ornn-api/CHANGELOG.md and ornn-web/CHANGELOG.md

Community

• Questions / how-to → Discussions → Q&A
• Ideas / RFCs → Discussions → Ideas
• Show off your agent integration → Discussions → Show & Tell
• Bug or feature → open an issue
• Security report → Private Vulnerability Reporting — see SECURITY.md
• Support guide → SUPPORT.md

Contributing

Pull requests are welcome. Before opening one, read CONTRIBUTING.md — it covers the issue-first workflow, branching, commit decomposition, and the changeset rule (CI blocks PRs without one).

By participating you agree to follow our Code of Conduct.

License

Apache License 2.0